The Modelled Year and the Experienced Year: Pricing to USD 171 Billion in a USD 46 Billion Half

Verisk’s 2026 Global Modeled Catastrophe Losses Report, published on 1 September, raised the global insured average annual loss to USD 171 billion, up USD 19 billion in a year and the highest figure it has ever published, with USD 117 billion or 68 percent attributed to the United States and a rise from USD 59 billion when the series began in 2012. The market is pricing against an entirely different number: five consecutive quarters have now passed without a single insured catastrophe event above USD 10 billion, first-half insured losses came in at USD 46 billion against a ten-year average of USD 64 billion, and the property catastrophe rate index has fallen 6.6 percent at January 2025, 12 percent at January 2026 and roughly 16 percent more through the June and July renewals. Howden Re’s September pre-renewal report finds the sector’s economic value added spread near break-even by mid-2026 while reported returns on equity remained comfortably positive.

There are two catastrophe numbers in circulation this month and they point in opposite directions. One is the loss the industry should expect in an average year. The other is the loss the industry actually sustained. The first has never been higher. The second has rarely been lower. Price is following the second, and that is the whole of the technical problem facing the January renewal.

The number the models produce

Verisk published its 2026 Global Modeled Catastrophe Losses Report on 1 September and raised the global insured average annual loss to USD 171 billion, an increase of USD 19 billion in twelve months and the highest figure in the history of the series. Of that total, USD 117 billion, or 68 percent, is attributed to the United States. The trajectory is the more instructive part: when Verisk first published the measure in 2012 it stood at USD 59 billion, and the near-tripling since reflects both expanded model coverage across more than twenty new countries and regions and genuine advances in the science behind the models. Verisk is careful to state what the figure is and is not. It is a modelled long-term benchmark derived from simulation across perils and geographies. It is not a forecast of 2026, or of any single year. That distinction is the one the market is currently failing to hold.

The number the market experienced

Against a modelled expectation of USD 171 billion, the experienced result has been exceptionally mild. Five consecutive quarters have now passed without a single insured catastrophe event exceeding USD 10 billion. Global insured natural catastrophe losses reached USD 46 billion in the first six months of 2026, the lowest first-half figure since 2019 and some 28 percent below the ten-year first-half average of USD 64 billion, with Swiss Re Institute’s preliminary estimate of USD 42 billion sitting well under its own trend figure of USD 66 billion. The rate response has been mechanical. The benchmark global property catastrophe index fell 6.6 percent at January 2025, a further 12 percent at January 2026, and roughly 16 percent more through the June and July 2026 renewals. Nor is relief expected from the weather: the National Oceanic and Atmospheric Administration places a 55 percent probability on a below-normal Atlantic hurricane season as a robust El Nino builds, against 35 percent for near-normal and 10 percent for above-normal. The temptation such a run creates is to treat five quiet quarters as evidence about the underlying distribution. They are not. They are five observations from the left tail of a distribution whose modelled mean has just been revised upward.

Accounting profit, economic break-even

Howden Re’s September pre-renewal report, Breaking the Glass, supplies the measurement that settles the argument. The economic value added spread for the global sector was near break-even by mid-2026 even as reported returns on equity stayed positive and accounting metrics looked healthy. Profitability is strong, capital is abundant, pricing continues to soften, and none of that indicates a less risky world. The report adds a structural observation of direct relevance to any reinsurer assessing where it sits in the chain: reinsurers paid less than 25 percent of all natural catastrophe losses in 2025, the fourth consecutive year below that threshold, because the higher attachment points established since January 2023 have pushed retention back onto cedants. Howden Re concludes that a substantial deterioration in underwriting or financial conditions would now be required to reverse the softening momentum. Read plainly, the sector is generating accounting returns while creating little or no economic value, and it is doing so in a year in which the weather did it a favour. That is the precise configuration in which capital is quietly consumed rather than compounded.

Latin America: the same question in our classes

The region imports this logic without importing the discipline that should accompany it. AM Best’s September market segment report describes a Latin American market that remains firmly favourable to cedants, with ample capacity, flexible terms and facultative pricing reductions of between 5 and 20 percent in some areas, set against regional economic growth revised down to 2.2 percent and a protection gap in which less than 24 percent of almost USD 21 billion in economic losses was insured. Fitch holds a neutral 2026 view on every rated Latin American insurance market except Mexico, where the value added tax reform has deteriorated the outlook. The uncomfortable part is that our classes have no equivalent of a published catastrophe index to anchor the debate. In Group Life and Personal Accident the analogue of an average annual loss is a burning cost normalised for exposure, trend and scheme composition, and that figure is produced internally or not at all. A market that will not price to a modelled benchmark when one is published and audited will certainly not price to one it has to construct for itself.

What this means for Power Re

Power Re underwrites Group Life and Personal Accident across Latin America, and the discipline that follows from a USD 171 billion expectation met by a USD 46 billion half is a pricing discipline, stated without qualification. Four commitments follow. First, we price to the modelled expectation, never to the experienced year. Every treaty carries an explicit expected loss built from exposure, trend and normalised burning cost, and a quiet period reduces the uncertainty load around that expectation only to the extent the additional data statistically warrants. It never reduces the expectation itself. Second, benign weather is not evidence about our classes. Mortality and morbidity accumulation in Group Life and Personal Accident responds to pandemic, transport, industrial and crowd events, and five quarters without a large property catastrophe carries no information whatsoever about the tail we actually carry. Our aggregation control states, before any event, how much exposure sits inside one geography, one cedant scheme and one event definition. Third, we measure our result the way Howden Re measures the sector, against the cost of the capital employed rather than against zero. A treaty that earns an accounting profit while returning less than the capital it consumes has destroyed value with a positive number attached to it, and we would rather identify that ourselves than have an agency identify it for us. Fourth, a softening market rewards the carrier willing to write the business others have correctly declined, and that reward is collected in the same year and repaid over the following five. Power Re does not grow by volume. It grows by quality, consistency and risk-adjusted return, and the region’s protection gap will be closed by the reinsurers that priced this cycle to the distribution rather than to the sample, and were still solvent, credible and rated when the demand finally arrived.